Update browser actions api to be like new design doc.

Update browser actions api to be like new design doc.

BUG=23879
TEST=Install sample gmail browser action sample.

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=29068

[Matt Perry, 2009-10-13 18:26:12]

LGTM. I guess we're OK with a breaking change to browser actions?

http://codereview.chromium.org/264046/diff/3002/4010
File chrome/renderer/resources/extension_process_bindings.js (right):

http://codereview.chromium.org/264046/diff/3002/4010#newcode344
Line 344: console.log(details);
stray debug statement

[Matt Perry, 2009-10-13 23:07:01]

http://codereview.chromium.org/264046/diff/5002/6009
File chrome/common/extensions/api/extension_api.json (right):

http://codereview.chromium.org/264046/diff/5002/6009#newcode852
Line 852: "imageData": {
Note: removing the property validation here will give people cryptic error
messages and/or crash the renderer if they pass an imageData object that is not
really a canvas ImageData.

[Aaron Boodman, 2009-10-13 23:29:20]

http://codereview.chromium.org/264046/diff/5002/6009
File chrome/common/extensions/api/extension_api.json (right):

http://codereview.chromium.org/264046/diff/5002/6009#newcode852
Line 852: "imageData": {
On 2009/10/13 23:07:02, Matt Perry wrote:
> Note: removing the property validation here will give people cryptic error
> messages and/or crash the renderer if they pass an imageData object that is
not
> really a canvas ImageData.

I see. I removed the width/height because it wasn't a perfect test (doesn't
really mean the thing is an image data) and because it seemed like it would be
misleading in the doc.

Can we be more defensive handling that api call instead?

[Matt Perry, 2009-10-13 23:33:36]

On 2009/10/13 23:29:20, Aaron Boodman wrote:
> http://codereview.chromium.org/264046/diff/5002/6009
> File chrome/common/extensions/api/extension_api.json (right):
> 
> http://codereview.chromium.org/264046/diff/5002/6009#newcode852
> Line 852: "imageData": {
> On 2009/10/13 23:07:02, Matt Perry wrote:
> > Note: removing the property validation here will give people cryptic error
> > messages and/or crash the renderer if they pass an imageData object that is
> not
> > really a canvas ImageData.
> 
> I see. I removed the width/height because it wasn't a perfect test (doesn't
> really mean the thing is an image data) and because it seemed like it would be
> misleading in the doc.
> 
> Can we be more defensive handling that api call instead?

We can do the tests elsewhere, but we can only really test that the object has
data, width, and height, right? So wherever we do it, we're just duplicating
what json schema would do.

[Aaron Boodman, 2009-10-14 01:54:31]

On 2009/10/13 23:33:36, Matt Perry wrote:
> On 2009/10/13 23:29:20, Aaron Boodman wrote:
> > http://codereview.chromium.org/264046/diff/5002/6009
> > File chrome/common/extensions/api/extension_api.json (right):
> > 
> > http://codereview.chromium.org/264046/diff/5002/6009#newcode852
> > Line 852: "imageData": {
> > On 2009/10/13 23:07:02, Matt Perry wrote:
> > > Note: removing the property validation here will give people cryptic error
> > > messages and/or crash the renderer if they pass an imageData object that
is
> > not
> > > really a canvas ImageData.
> > 
> > I see. I removed the width/height because it wasn't a perfect test (doesn't
> > really mean the thing is an image data) and because it seemed like it would
be
> > misleading in the doc.
> > 
> > Can we be more defensive handling that api call instead?
> 
> We can do the tests elsewhere, but we can only really test that the object has
> data, width, and height, right? So wherever we do it, we're just duplicating
> what json schema would do.

You're right, I'd just prefer to not have these properties show up in the doc.
So I did the check manually in JavaScript.

[Matt Perry, 2009-10-14 17:47:29]

LGTM